David lynch had the perfect response for this

I just published State Machines to Living Systems: Design Tenets for Intelligent Systems.

Reinforcement Learning also provides great philosophical insights. Agents that start off with an optimistic outlook of future rewards learn much better and much faster.

"The perfect tool or product always translates desires to reality immediately" - this hits different 5 years later now, when we have tools that can follow instructions in natural language to think, imagine and program for us. An unprecedented explosion in creativity is coming.

So my (sneaky) AI CLI stopped calling a few functions in the code. It did so after failing to fix an issue I raised multiple times which involved those functions. The behavior I cared about got fixed, but the functionality was completely broken. Vibe coders beware.

To summarize: 1. Make your AI plan and think before acting 2. Ensure its thinking is logged in a lot of detail for you to comment and modify (I might use comment threads on a Google doc next) 3. Multiple small changes implemented sequentially are better than a large batch of

Test failures at every milestone were minimal, and a lot of progress was made with minimal waste of resources. I ended the interaction by asking Gemini to summarise all instances where I provided feedback and what it learned from the modifications, and added it to a context file

Once we settled on the plan, every single step was executed one by one, where I pointed out unnecessary deletions, asked to use postgres instead of sqlite, asked for timezone awareness in very specific ways, and gently nudged on very small well defined points since they were

I wanted segregation of data fetching from processing, model building and all the downstream actions. I wanted better abstraction of the data. I wanted reorganisation, not rewrites. I outlined my goals in a markdown and asked specifically for a plan. Not code. An implementation

A whole day was spent. Close to a hundred iteration loops happened. There was progress for sure. I did have lots of data in my database. But it was all very poorly done. Future parts of the pipeline never worked with the database. Everything was discarded. Now let's talk about

Approach 1: A naive approach - told Gemini CLI to add a database layer to the code. It made smart-looking changes. I accepted them all, since it seemed to know what it was doing. Approach 2: A more systematic approach that involved careful planning before actual coding. This

The first task was to add a database layer to the code. The codebase fetched a lot of data through various data sources and processed the data before training a RL model. I wanted to persist the raw and processed data in a database so it's more useful. However, the data

After burning through a lot of API credits on my first few interactions with Claude Code, I was happy to see the free limits recently added to Gemini CLI (they are substantial) I'm working on a personal project which is a major refactor to a public codebase with the addition of

I performed the same coding task with the same model in two different ways. One way made me throw all the work away. The other created such great progress that I felt the urge to pat Gemini on the head, give it many treats and call it a good boy. This week in coding with LLMs:

Since AI is making it so much easier to build and maintain software, the resulting explosion in software would make distribution significantly tougher. Every single niche will have many many products all fighting for attention. Growth for new products will be 100x harder.

The bots at https://t.co/0jLp7734fS seem to all have issues and they throw it right at the user. I thought this thing was for humans to therapise with the AI, not the other way around. What even is this lol

Life is about learning something, then unlearning the same thing in a different context, and later wondering if you learned anything at all

When you start a startup, so many people tell you to ship fast and ship messy. "You’ll know you have PMF when people use your broken product." I think this is outdated advice that doesn't work anymore in 2024. Maybe better today would be: "ship fast, small scope, high quality"

Rewatched LOST just a few months back. I was among those who thought the last 2 seasons were bad on 1st viewing. The 2nd viewing was sublime. The issue was that I saw it as a flight crash story. This show wants to be about so much more. Better to just take it in as it is.

Terrace of a Café at Night' painting by Van Gogh, animated 📹 a_surotdinov https://t.co/xAyGqgfPi6